[PATCH 4/8] amd/amdkfd:fix overflowed constant in the function svm_migrate_copy_to_ram

Christian König christian.koenig at amd.com
Thu May 30 14:11:29 UTC 2024



Am 30.05.24 um 05:48 schrieb Jesse Zhang:
> If the svm migration copy memory gart fails or the dma mapping page fails for the first time.
> But the variable i is still 0, and executing i-- will overflow.
>
> Signed-off-by: Jesse Zhang <Jesse.Zhang at amd.com>
> ---
>   drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> index 8ee3d07ffbdf..3620eabf13c7 100644
> --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c
> @@ -650,9 +650,10 @@ svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange,
>   out_oom:
>   	if (r) {
>   		pr_debug("failed %d copy to ram\n", r);
> -		while (i--) {
> +		while (i) {
>   			svm_migrate_put_sys_page(dst[i]);
>   			migrate->dst[i] = 0;
> +			i--;

That looks incorrect to me.

"i" is usually the entry which failed and doesn't need to cleanup. So 
using "while (i---) ...." is a very common and correct way to clean 
things up.

With the code changed as above 0 for example would never be cleaned up.

Christian.

>   		}
>   	}
>   



More information about the amd-gfx mailing list